LLM7 (open-source free API gateway)

No sign-up needed for anonymous use — up to 500000 tokens/day within a rolling 24-hour window; register to get a free token and that rises to up to 1000000 tokens/day

The official TERMS.md (verified 2026-08-30) states verbatim: for anonymous use, 'Up to 500,000 tokens/day over a rolling 24-hour window, with up to 1 text request/s, 10 text requests/min, and 60 text requests/hour'; after registering, 'Up to 1,000,000 tokens/day over a rolling 24-hour window, with up to 2 text requests/s, 60 text requests/min, and 250 text requests/hour' — the quota is calculated as input plus output tokens combined. A live query against the public /v1/models endpoint returned 50 callable models, spanning chat (naming across GPT/Claude/Gemini/DeepSeek/GLM/Grok/Mistral/Llama/MiniMax-style families) plus image and video generation; the exact lineup shifts over time — check that endpoint for the current list. Where third-party channels report different quota numbers than the official TERMS.md cited here, defer to the latest official terms.

⚠️An open-source gateway — the GitHub repo chigwell/llm7.io is public and its maintainer's identity is known, and it publishes a formal TERMS.md — making it one of the few relay-type services that passed this round of verification (several 'New-API-panel reskin' gray-market relay sites in the same batch were rejected). That said, the official terms state outright that 'The Service may call, route to, experiment with, or otherwise host third-party models,' meaning requests may be routed to third-party models or infrastructure; this is an individually/community-maintained project, not an official offering from any model vendor — evaluate production reliability yourself.